Developer Documentation

Driverless Vehicles

The design document for phase 1 of incorporating driverless vehicles into VE is at https://github.com/gregorbj/OregonDOT-VisionEval/blob/master/documents/drafts/Driverless_Vehicle_Mods.md

Desired additions to driverless vehicle modeling

Following is a list of desired additions to driverless vehicle modeling. The first 2 items will be done in the next phase of work.

1. Enable non-drivers to become drivers with driverless cars (e.g., youth and eldery). Dial to change drivers rate in AssignDrivers module (or simply use region_hh_driver_adjust_prop.csv), and possibly other implications. Source for how much these populations would drive?
2. Occupancy dials for owned and CarService vehicles. Current model implicitly assumes NHTS average occupancy, allow user to change this value. How identify base value and source for reasonable alternative occupancy range?
3. Inputs limiting “driverless” to selected powertrains. This might be an easy non-combustion engine toggle, or expanding the planned 2 value input to 6.
4. Add a driverless vehicle maintenance (per mile) operating cost surcharge. This would reflect the cost to keep the software and AV technology in top condition. Source for additional costs?
5. Tabulate/enable output measure of overall Car Service impact to lend realism. Identify total and share of car service miles, number of households using Car Service (currently able?) and/or the number of CarService vehicles (i.e., for parking demands)
6. Car Service rate structure – add fixed rate to per mile rate.
7. Allow policy that taxes zero-passenger vehicle miles (owned or CarService). Would require separating out base amount of zero-passenger miles, and an ability to substitute between zero and non-zero miles. For car services, zero-passenger miles are miles traveled without a customer (same as deadhead miles). For owned vehicles, zero-passenger miles are miles traveled without any occupants.
8. Expand the revenue generation reporting from the model. Sources and amounts. (could do much with existing outputs, might also need to save intermediate outputs)
9. (Tara added 9/8) Replace random assignment of which HHs get AVs. Current input is X% of vehicles are AV in model year XXXX. Borrow HH index trick from PAYD module, where HH attributes weight the likelihood of the AV assignment. (e.g., given input that 5% vehicles made in 2030 are assumed AV, index gives higher probability that AVs go to EV vehicles of that year and HHs with teenagers, and longer commutes, exact attributes in index determined by lit review/survey data).
10. (Alex added 9/8) Be able to consider (model) varying shared price schemes such as:

• The standard we currently think of is a per mile charge, something similar to the way a TNC works today.
• Some areas have been playing with a public-private shared model, where TNCs help to supplement or replace transit. Playing that into the future, I could see scenarios where travelers for a region don’t pay for shared rides per mile, they would pay per ride. So regardless of distances (within a given region), they might get the option at a flat / subsidized fare.
• Americans hate to be constrained; they love the freedom to make their own choices. Because of this, companies have been able to capitalize on charging customers more for “unlimited” services. Example – American’s love free refills, even if the initial glass is $3. American’s love unlimited data plans, even if they typically only use X GBs per month. It’s not a stretch in my mind to envision a company offering a travel anywhere / anytime plan, especially if that company isn’t paying for driver labor, is advertising to the traveler, and is gathering important preference and marketing data about the traveler they are servicing. Anyway, I could imagine a scenario where travelers buy into unlimited miles plans and no longer see a per mile cost – their per mile costs go to zero all the cost might be as a sunk subscription cost (just like ownership, but where insurance, gas, maintenance… are all sunk costs).